Radiomics-based detection of acute myocardial infarction on noncontrast enhanced midventricular short-axis cine CMR images.

Cardiac magnetic resonance cine images are primarily used to evaluate functional consequences, whereas limited information is extracted from the noncontrast pixel-wise myocardial signal intensity pattern. In this study we want to assess whether characterizing this inherent contrast pattern of noncontrast-enhanced short axis (SAX) cine images via radiomics is sufficient to distinguish subjects with acute myocardial infarction (AMI) from controls. Cine balanced steady-state free-precession images acquired at 1.5 T from 99 AMI and 49 control patients were included. First, radiomic feature extraction of the left ventricular myocardium of end-diastolic (ED) and end-systolic (ES) frames was performed based on automated (AUTO) or manually corrected (MAN) segmentations. Next, top features were selected based on optimal classification results using a support vector machine (SVM) approach. The classification performances of the four radiomics models (using AUTO or MAN segmented ED or ES images), were measured by AUC, classification accuracy (CA), F1-score, sensitivity and specificity. The most accurate model was found when combining the features RunLengthNonUniformity, ClusterShade and Median obtained from the manually segmented ES images (CA = 0.846, F1 score = 0.847). ED analysis performed worse than ES, with lower CA and F1 scores (0.769 and 0.770, respectively). Manual correction of automated contours resulted in similar model features as the automated segmentations and did not improve classification results. A radiomics analysis can capture the inherent contrast in noncontrast mid-ventricular SAX cine images to distinguishing AMI from healthy subjects. The ES radiomics model was more accurate than the ED model. Manual correction of the autosegmentation did not provide significant classification improvements.

Rapid dynamic R1 /R2 */temperature assessment: a method with potential for monitoring drug delivery.

Local drug delivery by hyperthermia-induced drug release from thermosensitive liposomes (TSLs) may reduce the systemic toxicity of chemotherapy, whilst maintaining or increasing its efficacy. Relaxivity contrast agents can be co-encapsulated with the drug to allow the visualization of the presence of liposomes, by means of R2 *, as well as the co-release of the contrast agent and the drug, by means of R1, on heating. Here, the mathematical method used to extract both R2 * and R1 from a fast dynamic multi-echo spoiled gradient echo (ME-SPGR) is presented and analyzed. Finally, this method is used to monitor such release events. R2 * was obtained from a fit to the ME-SPGR data. Absolute R1 was calculated from the signal magnitude changes corrected for the apparent proton density changes and a baseline Look-Locker R1 map. The method was used to monitor nearly homogeneous water bath heating and local focused ultrasound heating of muscle tissue, and to visualize the release of a gadolinium chelate from TSLs in vitro. R2 *, R1 and temperature maps were measured with a 5-s temporal resolution. Both R2 *and R1 measured were found to change with temperature. The dynamic R1 measurements after heating agreed with the Look-Locker R1 values if changes in equilibrium magnetization with temperature were considered. Release of gadolinium from TSLs was detected by an R1 increase near the phase transition temperature, as well as a shallow R2 * increase. Simultaneous temperature, R2 * and R1 mapping is feasible in real time and has the potential for use in image-guided drug delivery studies.

MRI contrast variation of thermosensitive magnetoliposomes triggered by focused ultrasound: a tool for image-guided local drug delivery.

Improved drug delivery control during chemotherapy has the potential to increase the therapeutic index. MRI contrast agent such as iron oxide nanoparticles can be co-encapsulated with drugs in nanocarrier liposomes allowing their tracking and/or visualization by MRI. Furthermore, the combination of a thermosensitive liposomal formulation with an external source of heat such as high intensity focused ultrasound guided by MR temperature mapping allows the controlled local release of the content of the liposome. MRI-guided high-intensity focused ultrasound (HIFU), in combination represents a noninvasive technique to generate local hyperthermia for drug release. In this study we used ultrasmall superparamagnetic iron oxide nanoparticles (USPIO) encapsulated in thermosensitive liposomes to obtain thermosensitive magnetoliposomes (TSM). The transverse and longitudinal relaxivities of this MRI contrast agent were measured upon TSM membrane phase transition in vitro using a water bath or HIFU. The results showed significant differences for MRI signal enhancement and relaxivities before and after heating, which were absent for nonthermosensitive liposomes and free nanoparticles used as controls. Thus, incorporation of USPIO as MRI contrast agents into thermosensitive liposomes should, besides TSM tumor accumulation monitoring, allow the visualization of TSM membrane phase transition upon temperature elevation. In conclusion, HIFU under MR image guidance in combination with USPIO-loaded thermosensitive liposomes as drug delivery system has the potential for a better control of drug delivery and to increase the drug therapeutic index.

Carotid plaque high-resolution MRI at 3 T: evaluation of a new imaging score for symptomatic plaque assessment.

PURPOSE: To assess the sensitivity and specificity of intra-plaque hemorrhage (IPH), large lipid-rich necrotic core (LR-NC) and ulceration or cap rupture (UCR) for symptomatic carotid plaque characterization and to evaluate a new imaging score [Hemorrhage, Ulceration or cap rupture, Lipid-rich necrotic Core (HULC) score based on the sum of presence/absence of IPH, UCR and LR-NC; range 0-3] for assessment of recently symptomatic carotid plaques. MATERIAL AND METHODS: Twenty-seven recently symptomatic (<8 weeks) and 36 asymptomatic patients with a carotid plaque thicker than 2 mm were prospectively imaged on a 3-T magnetic resonance (MR) system using high-resolution, multi-contrast MR sequences. Prior to analysis, all images were reviewed to assess image quality of each sequence. Sensitivity and specificity of IPH, LR-NC, UCR and HULC scores were calculated. RESULTS: Fifty-one patients were analyzed (26 symptomatic carotids and 67 asymptomatic carotids) after exclusion of studies with poor image quality. Sensitivity and specificity for symptomatic carotid plaque was, respectively, 46.1% and 97% for IPH, 84.6% and 73.1% for UCR and 80.7% and 76.1% for LR-NC. A HULC score of 2 or more showed a sensitivity of 73% and a specificity of 92.5%. CONCLUSION: At 3 T, intra-plaque hemorrhage is the most specific criterion to characterize symptomatic carotid plaque. The HULC score offers the best compromise between sensitivity and specificity.

4D time-resolved magnetic resonance angiography for noninvasive assessment of pulmonary arteriovenous malformations patency.

PURPOSE: To assess the capability of four-dimensional (4D) time-resolved magnetic resonance angiography (MRA) to assess pulmonary arteriovenous malformations (PAVMs) patency by analyzing pulmonary arterial and venous enhancement kinetics. MATERIALS AND METHODS: Seven patients with eight documented patent PAVMs underwent a 4D-MRA with keyhole and viewsharing compression at 3T with the following parameters: spatial resolution 0.87 × 0.87 × 1.4 mm(3); field of view 500 × 350 × 238 mm(3); dynamic scan time (temporal resolution) 1.2 seconds; total acquisition time 18.1 seconds for six dynamic datasets (6 × 1.2 sec + reference scan: 10.9 sec). All images were reviewed by two experienced radiologists. Image quality was rated on a qualitative 5-point scale (1: not assessable to 5: excellent). Signal value was measured on cross-sectional planes for the afferent arteries and efferent veins of the PAVM, and for normal reference healthy arteries and veins. The difference in time to peak for each coupled artery/vein (dTTPav) was calculated and compared with a Mann-Whitney test between PAVMs and reference vessels. RESULTS: Mean image quality was 3.2 ± 0.9. dTTPav was significantly smaller in PAVMs (0.15 ± 0.76 sec) than in reference vessels (3.75 ± 1.62 sec), P < 0.001. CONCLUSION: 4D-MRA is a promising tool for noninvasive assessment of PAVM patency.

Radiofrequency ablation of small liver malignancies under magnetic resonance guidance: progress in targeting and preliminary observations with temperature monitoring.

OBJECTIVES: To evaluate the feasibility and effectiveness of magnetic resonance (MR)-guided radiofrequency (RF) ablation for small liver tumours with poor conspicuity on both contrast-enhanced ultrasonography (US) and computed tomography (CT), using fast navigation and temperature monitoring. METHODS: Sixteen malignant liver nodules (long-axis diameter, 0.6-2.4 cm) were treated with multipolar RF ablation on a 1.5-T wide-bore MR system in ten patients. Targeting was performed interactively, using a fast steady-state free precession sequence. Real-time MR-based temperature mapping was performed, using gradient echo-echo planar imaging (GRE-EPI) and hardware filtering. MR-specific treatment data were recorded. The mean follow-up time was 19 +/- 7 months. RESULTS: Correct placement of RF electrodes was obtained in all procedures (image update, <500 ms; mean targeting time, 21 +/- 11 min). MR thermometry was available for 14 of 16 nodules (88%) with an accuracy of 1.6 degrees C in a non-heated region. No correlation was found between the size of the lethal thermal dose and the ablation zone at follow-up imaging. The primary and secondary effectiveness rates were 100% and 91%, respectively. CONCLUSIONS: RF ablation of small liver tumours can be planned, targeted, monitored and controlled with MR imaging within acceptable procedure times. Temperature mapping is technically feasible, but the clinical benefit remains to be proven.

SNR enhancement of highly-accelerated real-time cardiac MRI acquisitions based on non-local means algorithm.

Real-time cardiac MRI appears as a promising technique to evaluate the mechanical function of the heart. However, ultra-fast MRI acquisitions come with an important signal-to-noise ratio (SNR) penalty, which drastically reduces the image quality. Hence, a real-time denoising approach would be desirable for SNR amelioration. In the clinical context of cardiac dysfunction assessment, long acquisitions are required and for most patients the acquisition takes place with free breathing. Hence, it is necessary to compensate respiratory motion in real-time. In this article, a real-time and interactive method for sequential registration and denoising of real-time MR cardiac images is presented. The method has been experimented on 60 fast MRI acquisitions in five healthy volunteers and five patients. These experiments assessed the feasibility of the method in a real-time context.

Validation of fast MR thermometry at 1.5 T with gradient-echo echo planar imaging sequences: phantom and clinical feasibility studies.

The purpose of this work was to validate in phantom studies and demonstrate the clinical feasibility of MR proton resonance frequency thermometry at 1.5 T with segmented gradient-echo echo planar imaging (GRE-EPI) sequences during liver tumour radiofrequency (RF) ablation. Classical GRE acquisitions and segmented GRE-EPI acquisitions were performed at 1.5 T during simultaneous RF heating with an MR-compatible RF electrode placed in an agar gel phantom. Temperature increments were calculated and compared with four optical temperature probe measurements using Bland- Altman analysis. In a preliminary clinical feasibility study, the rapid GRE-EPI sequence (echo train length = 13) was used for MR temperature monitoring of RF ablation of liver tumours in three patient procedures. For phantom experiments, the Bland-Altman mean of differences between MR and optical probe temperature measurements was <0.4 degrees C, and the 95% limits of agreement value was <1.4 degrees C. For the in vivo studies, respiratory-triggered GRE-EPI acquisitions yielded a temperature accuracy of 1.3 +/- 0.4 degrees C (acquisition time = 0.6 s/image, spatial coverage of three slices/respiratory cycle). MR proton resonance frequency thermometry at 1.5 T yields precise and accurate measurements of temperature increment with both classical GRE and rapid GRE-EPI sequences. Rapid GRE-EPI sequences minimize intra-scan motion effects and can be used for MR thermometry during RF ablation in moving organs.

New horizons in MR-controlled and monitored radiofrequency ablation of liver tumours.

There is a sustained interest in using magnetic resonance (MR) thermometry to monitor the radiofrequency ablation of liver tumours as a means of visualizing the progress of the thermal coagulation and deciding the optimal end-point. Despite numerous technical challenges, important progress has been made and demonstrated in animal studies. In addition to MR thermometry, MR can now be used for the guidance of the tumour targeting with 'fluoroscopic' rapid image acquisition, and it can provide several contrast mechanisms for post-procedural assessment of the extent of the thermal coagulation zone. Challenges of in vivo simultaneous MR thermometry implementation and the current limitations of the thermal dose model for the estimation of the extent of the thermal coagulation zone are discussed. MR imaging could enhance the success of RF ablation of liver tumours due to its potential to provide accurate targeting, monitoring, and post-procedural evaluation.

Theory-based signal calibration with single-point T1 measurements for first-pass quantitative perfusion MRI studies.

RATIONALE AND OBJECTIVES: The aim of the study is to develop a theory-based signal calibration approach to be used for the conversion of signal-time curves to absolute contrast concentration-time curves for first-pass contrast-enhanced quantitative myocardial perfusion studies. MATERIALS AND METHODS: A normalization procedure was used to obtain a theoretical relationship between image signal and T1 and perform rapid single-point T1 measurements. T1 measurements were compared with reference T1 measurements. The method also was used in preliminary in vivo contrast-enhanced first-pass perfusion studies, and its applicability for dual-delay-time acquisitions was shown. A theory-based error sensitivity analysis was used to characterize the robustness of the method. RESULTS: The normalization procedure was implemented with minimal noise enhancement and insensitivity to small misregistrations through postprocessing techniques. The rapid T1 measurements are in excellent agreement with the reference measurements (R = 0.99, slope = 1.05, bias = -5.96 milliseconds). For in vivo studies, it is possible to simultaneously calibrate the arterial input function and myocardial enhancement curves acquired with different effective trigger delays through appropriate use of the theory-based signal calibration model. With this method, errors of in vivo baseline T1 estimates are large, but the effect of these large errors on the accuracy of contrast agent concentration estimates is limited. CONCLUSION: This theory-based signal calibration approach can be used to perform rapid T1 mapping and provides flexibility for in vivo calibration of signal-time curves resulting from dual-delay-time first-pass contrast-enhanced acquisitions.

B(0) and B(1)-insensitive uniform T(1)-weighting for quantitative, first-pass myocardial perfusion magnetic resonance imaging.

Myocardial perfusion can be estimated, in principle, from first-pass MR images by converting the T(1)-weighted signal-time curves to contrast agent concentration-time curves. Typically, T(1) weighting is achieved by saturating the magnetization with a nonselective radiofrequency (RF) pulse prior to the imaging sequence. The accuracy of the perfusion estimate derived from the single-point T(1)-weighted signal depends on the initial residual longitudinal magnetization (RLM) produced by the saturation pulse. In this study we demonstrate that single-shot, echo-planar imaging can be used to show initial RLM resulting from incomplete saturation due to static magnetic field and RF field inhomogeneities in the heart at 1.5 T. Three saturation pulses, single, composite simple, and composite B(1)-insensitive rotation (BIR-4) were evaluated in phantom and cardiac experiments. The RLM image was calculated by normalizing the saturated image by a proton-density-weighted image. Mean RLM produced by the three saturation pulses was significantly different in noncontrast cardiac imaging (RLM(single) = 0.108 +/- 0.078; RLM(composite) = 0.051 +/- 0.052; RLM(BIR-4) = 0.011 +/- 0.009; P < 0.001; n = 20). Using a BIR-4 pulse to perform saturation of magnetization seems promising for improving the effectiveness and uniformity of T(1) weighting for first-pass perfusion imaging.